2.6.37: *ERROR* Attempting to mmap a purgeable buffer

2.6.37: *ERROR* Attempting to mmap a purgeable buffer

[Carlos R. Mafra]

I'd like to report an issue which just happened while using 2.6.37.

I was transfering some large files between two external harddisks
using btrfs and suddenly my window manager exited and I was left
in the login screen of kdm.

The end of dmesg looks like:

[42460.335771]  sdd: sdd1 sdd2 sdd3 sdd4
[42460.337496] sd 34:0:0:0: [sdd] Assuming drive cache: write through
[42460.337500] sd 34:0:0:0: [sdd] Attached SCSI disk
[42468.912626] device fsid 4b41009f616ab3b8-a8cf4357d5b54bb4 devid 1 transid 950 /dev/sdd2
[42468.913839] btrfs: use compression
[42893.878855] [drm:i915_gem_mmap_gtt_ioctl] *ERROR* Attempting to mmap a purgeable buffer

and before that there is nothing out of the usual. After starting the
window manager again all things look ok. I am reporting this anyway, in case
it might help improving the kernel and making it stabler.

The error string is from ab18282d58c ("drm/i915: Warn before mmaping a purgeable 
buffer.") from 2009, so I'm cc:ing the author.

My laptop is a 64-bit Vaio with

Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c)

and today was the first time this error happened, afaict.

Is there something else I can do to help? Is this something fixable in the
kernel or was it an error with my hardware? Suddenly being kicked out of X is 
quite bad, so I hope this can be fixed :-)

Re: 2.6.37: *ERROR* Attempting to mmap a purgeable buffer

[Chris Wilson]

On Sun, 9 Jan 2011 13:44:23 +0100, "Carlos R. Mafra" <crmafra2@gmail.com> wrote:
> Is there something else I can do to help? Is this something fixable in the
> kernel or was it an error with my hardware? Suddenly being kicked out of X is 
> quite bad, so I hope this can be fixed :-)

It's a use after free bug in the ddx, presumably. I haven't worked out how
yet since all active buffers should be appropriately marked. So either
there is a race in marking the buffer, or we failed to handle an error
appropriately or there is a real use-after-free. The kernel is being
gracious enough to warn you why userspace just crashed.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre